]> www.infradead.org Git - users/jedix/linux-maple.git/commitdiff
virtio-net: drop NETIF_F_FRAGLIST
authorJason Wang <jasowang@redhat.com>
Tue, 4 Aug 2015 18:34:04 +0000 (02:34 +0800)
committerSantosh Shilimkar <santosh.shilimkar@oracle.com>
Fri, 13 Nov 2015 17:23:50 +0000 (09:23 -0800)
virtio-net: drop NETIF_F_FRAGLIST

Orabug: 22154074
CVE-2015-5156

virtio declares support for NETIF_F_FRAGLIST, but assumes
that there are at most MAX_SKB_FRAGS + 2 fragments which isn't
always true with a fraglist.

A longer fraglist in the skb will make the call to skb_to_sgvec overflow
the sg array, leading to memory corruption.

Drop NETIF_F_FRAGLIST so we only get what we can handle.

Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 48900cb6af4282fa0fb6ff4d72a81aa3dadb5c39)
Signed-off-by: Brian Maly <brian.maly@oracle.com>
drivers/net/virtio_net.c

index 63c7810e1545a357eda7578af862ed18322de933..6ac04e2bff07588cd09c3c66a8af3f35ae6acdca 100644 (file)
@@ -1756,9 +1756,9 @@ static int virtnet_probe(struct virtio_device *vdev)
        /* Do we support "hardware" checksums? */
        if (virtio_has_feature(vdev, VIRTIO_NET_F_CSUM)) {
                /* This opens up the world of extra features. */
-               dev->hw_features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
+               dev->hw_features |= NETIF_F_HW_CSUM | NETIF_F_SG;
                if (csum)
-                       dev->features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
+                       dev->features |= NETIF_F_HW_CSUM | NETIF_F_SG;
 
                if (virtio_has_feature(vdev, VIRTIO_NET_F_GSO)) {
                        dev->hw_features |= NETIF_F_TSO | NETIF_F_UFO